| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Full paper |
Medical Physics and Clinical Engineering, City Hospital Campus, Nottingham University Hospitals NHS Trust, Hucknall Road, Nottingham NG5 1PB, UK
 |
Abstract |
|---|
 Top Abstract IntroductionMethodResultsDiscussionConclusionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionConclusionReferences |
|---|
Dose–area product (DAP) alone is not an adequate indicator of skin dose, since in many cases there is a poor correlation between the two quantities [8–12]. Slow radiographic film can be used to measure the dose distribution across the patient's skin. However, its useful range is limited by its saturation point. Kodak EDR2 film (Eastman Kodak Company, Rochester, NY) has previously been used for skin dosimetry [12–14], and is the least sensitive of the slow films designed for portal imaging and quality control in radiotherapy. However, it saturates at around 1 Gy to 1.5 Gy, depending on the processing conditions applied [13–15]. There is now a growing range of "Gafchromic" films (International Specialty Products, Wayne, NJ), which saturate at higher radiation doses and do not require processing, but at present these are much more expensive than those in the Kodak range. Smaller detectors such as thermoluminescent dosemeters and scintillation detectors cover only a small area, so are liable to miss the region of maximum dose.
In modern cardiac catheterization laboratories, information about the exposure parameters is stored in the DICOM file for each image series. The DICOM standard [16] specifies fields for data such as the number of frames in the series, angulation of the X-ray imaging unit, detector position and field size, imaging mode, beam energy and tube current. If these fields are populated, they allow the position and magnitude of the radiation dose to the patient's skin to be estimated, for each image series. Since fluoroscopic images are not usually stored, this detailed dose information is only available for the acquisition runs.
The purpose of this study was to develop a mathematical model to calculate the skin dose distribution across the patient's back, using the exposure and projection data stored in the image files. Maximum skin doses predicted by the model were compared with film dosimetry measurements for coronary angiography (CA) and percutaneous transluminal coronary angioplasty (PTCA) procedures. Three methods for including an estimated contribution from fluoroscopy were investigated.
|
Method |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionConclusionReferences |
|---|
The dose model
The exposure information is first extracted from all of the image files in the patient examination folder, and written to a structure array. An example of these data is shown in Table 1
. Each row relates to one file, i.e. one image series. The columns contain the series or run number (Run), number of frames in that series (Frames), imaging protocol (Protocol), peak beam kilovoltage (kVp), tube current (mA), pulse width (ms), primary and secondary angles describing the orientation of the imaging unit in degrees (Ang1 and Ang2), source to image distance in millimetres (SID) and detector field size in millimetres (II).
|
Additional data required by the model are:
|
|
The entrance dose rate to the phantom was measured using a Radcal 9010 series dosemeter, with a 60 cm3 ionization chamber (Radcal, Monrovia, CA), as shown in Figure 1. The standard clinical acquisition mode (12.5 FPS Coronary) was used. This mode does not employ any copper filtration.
|
The dose distribution is calculated at 1 mm intervals in a plane at the height of the couch top, that extends from –30 cm to +30 cm in the x (cross-couch) direction and –20 cm to +20 cm in the y (parallel to couch axis) direction. Positive x is defined towards the right-hand side of the patient, and positive y towards the patient's head, so that the dose map appears as though viewing the patient from behind.
For each acquisition run, the position of the focal spot (xspot, yspot, zspot) is calculated in Cartesian coordinates, using Equations (2) to (4).
|
|
|
|
|
|
|
|
Beam limitation is calculated assuming square radiation fields, and ignoring any secondary collimation applied by the user. The angle (
) between the beam central axis and its collimated outer edge is calculated from the field size and SID.
"SIDcoll" is the radius of a sphere centred at the focal spot that passes through the four corners of the collimated field, at the detector face. It is calculated from field size and SID, using Pythagoras' law. The positions of the field corners at the detector face are calculated in Cartesian coordinates, using Equations (6) to (8).
|
|
|
|
|
|
|
|
|
|
The completed dose maps from each run are summed to give a total dose array, which is then displayed as a filled contour map.
The model described above considers only the acquisition run data, which is stored in the image files. In clinical practice, some examinations involve extended fluoroscopy times, and where fluoroscopy is performed primarily in one projection this can raise the maximum skin dose considerably. Three options for estimating and including the contribution from fluoroscopy were investigated.
Comparison of calculated and measured doses
A dose survey has previously been carried out in our cardiac catheterization laboratory, using Kodak EDR2 film to measure the skin dose distribution across the patient's back [12]. The current study employed the dose model to calculate a dose map for each of the patients included in the survey. The maximum doses determined by the two methods were compared.
The calculated skin dose maps were first compared visually with the films. Patients were excluded from the study if the region of maximum dose predicted by the model was outside the area of the film.
For the subset of films that showed no saturation, the Pearson correlation coefficients between calculated and measured maximum doses were computed. This subset included 14 CA and 20 PTCA procedures.
|
Results |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionConclusionReferences |
|---|
shows an example of the data output from the skin dose model. A visual comparison with the corresponding dosimetry film (Figure 3) shows that the region of maximum dose has been correctly identified. The fields visible on the calculated dose map can be broadly matched with some of those appearing on the film. However, fields arising from purely fluoroscopic exposures are seen only on the film, since the model has no projection data for fluoroscopy.
|
|
shows the calculated doses for procedures resulting in film saturation, for each version of the dose model. The film saturates at about 1 Gy, so saturation indicates a skin dose of 1 Gy or more [15].
|
On applying any of the correction methods to include the contribution from fluoroscopy, the calculated dose to this patient increased to more than 1 Gy. When corrected by fluoroscopy time or concentration factor, the model successfully identified all procedures that resulted in film saturation. When adjusted by the ratio of displayed to calculated DAP, the model predicted one of these patients to have a maximum skin dose of 928 mGy, and all others to have maximum doses of more than 1 Gy.
Figures 5–8 show calculated versus measured doses for all the procedures that did not result in film saturation, i.e. for which measured doses were less than 1 Gy. The error bars show the expected uncertainty in film dosimetry measurements [12]. In each case, a trend line has been fitted, that passes through the origin. The equation of the trend line and the square of the Pearson correlation coefficient are shown on each figure.
|
|
|
|
), suggesting that the model underestimates doses by 26% on average. There was a strong correlation between calculated and measured doses (R2 = 0.672).
On using DAP or fluoroscopy time to estimate the contribution from fluoroscopy the gradient of the trend line was closer to 1 (Figures 6 and 7). The strength of the correlation also increased (R2 = 0.708 and R2 = 0.716, respectively).
The product of concentration factor and DAP was multiplied by 181 000, to force the gradient of the trend line to 1 (Figure 8). This method gave the strongest correlation between calculated and measured doses (R2 = 0.735).
|
Discussion |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionConclusionReferences |
|---|
Using only the acquisition data stored in the DICOM files can lead to large errors in dosimetry for examinations involving long fluoroscopy times and few projections. Three options for including an estimated contribution from fluoroscopy were explored. Each improved the accuracy of the model for procedures involving a large proportion of fluoroscopy, and predicted doses of at least 928 mGy for all procedures resulting in film saturation. The method using displayed DAP together with a "concentration factor" gave the strongest correlation between calculated and measured doses. Using this method, 95% of calculated doses were within ± 270 mGy of measured doses, for films showing no saturation.
Each method for including the contribution from fluoroscopy requires additional information (either DAP or fluoroscopy time) that is not stored in the image files and must therefore be obtained from another source. At present, displayed DAP and fluoroscopy time for all procedures are manually recorded in a book, and in an Oracle database (Oracle Corporation, Redwood Shores, CA). In the near future, these may be automatically stored in a new catheterization laboratory information system.
It is hoped that clinical application of the dose model can eventually be completely automated – to extract the relevant data from the image files and obtain the fluoroscopy time from the information system or database as each examination is completed, to calculate the skin dose distribution and alert a member of staff if the maximum dose exceeds a certain threshold. This would enable staff to follow up at-risk patients by examining their skin and warning them of potential effects before they leave the hospital.
The accuracy of the model is limited by a number of unknown variables, about which assumptions have had to be made. No exposure factors are stored for fluoroscopy, as has been previously discussed, necessitating an estimation of the contribution to total dose from the fluoroscopic parts of the procedure. The actual couch position for each procedure is not known, so it is necessary to assume a certain couch height, and to ignore any horizontal movement. No information about secondary collimation or the use of the wedge filter is available, so these must be assumed not to have been used.
Improved accuracy is dependent on manufacturers building in the facilities to make this information available. Whilst den Boer et al [17] and Chugh et al [18] have worked with equipment manufacturers to develop real-time skin dose monitoring software that utilizes much of this information, none is currently available for purchase.
As with film dosimetry, the model considers only those radiation beams that pass through the plane of the couch, and ignores any contributions to skin dose from lateral views. A potential improvement would be to use a three-dimensional model of the patient to estimate skin doses over the whole surface of the thorax. However, this would require couch positioning data, to achieve any degree of accuracy.
|
Conclusion |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionConclusionReferences |
|---|
The accuracy of the model is limited by several unknown variables that are not recorded by the imaging system. Improvement of dosimetric accuracy is dependent upon manufacturers developing methods for storing and accessing this information.
|
Acknowledgments |
|---|
Received for publication April 22, 2005. Revision received May 17, 2006. Accepted for publication June 20, 2006.
|
References |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionConclusionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  C. J. Martin RADIATION DOSIMETRY FOR DIAGNOSTIC MEDICAL EXPOSURES Radiat Prot Dosimetry, March 28, 2008; (2008) ncm495v1. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P P Dendy Radiation risks in interventional radiology Br. J. Radiol., January 1, 2008; 81(961): 1 - 7. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| BJR | DMFR | IMAGING | ALL BIR JOURNALS |
